[MAGNOLIA-2507] performance issues with JBoss and behavior of the cache Created: 11/Dec/08  Updated: 27/Nov/13  Resolved: 27/Nov/13

Status: Closed
Project: Magnolia
Component/s: cache
Affects Version/s: 3.6.1
Fix Version/s: None

Type: Bug Priority: Major
Reporter: Juan Carlos, Antonio Assignee: Unassigned
Resolution: Cannot Reproduce Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

server Linux with:
•Java = 1.5.0_12 de sun
•JBoss = 4.0.4.GA (usamos la instancia ALL)
•S.O. = Suse Linux Enterprise Server 9 SP3
•MySQL = 5.0.67-community
•Magnolia = CE 3.6.1


Attachments: File test.zip    
Template:
Acceptance criteria:
Empty
Task DoD:
[ ]* Doc/release notes changes? Comment present?
[ ]* Downstream builds green?
[ ]* Solution information and context easily available?
[ ]* Tests
[ ]* FixVersion filled and not yet released
[ ]  Architecture Decision Record (ADR)
Bug DoR:
[ ]* Steps to reproduce, expected, and actual results filled
[ ]* Affected version filled

 Description   

The 4 tests are as follows:

  • With Tomcat and templates without parameters (...rutaGaudi.ca_ES.html)
  • With Tomcat and templates with parameters (...rutaGaudi.html?leng=ca_ES)
  • With JBoss and templates without parameters (...rutaGaudi.ca_ES.html)
  • With JBoss and templates with parameters (...rutaGaudi.html?leng=ca_ES)

The Public's cache is set to come by default in the distribution with Tomcat.

We find the following in the different tests:

Try-TOMCAT and without parameters
1.in 360s update the page "ComViatjar" (page from another branch), and we have movement in the BD and delay in the English language alone.
2.in 600s update the page "RutaGaudi" (the test), and we are moving in the BD, and a slight delay in the 3 languages.

Try-JBoss and without parameters
1.in 360s update the page "ComViatjar" , and we have movement in the BD and significant delays.
2.in 600s update the page "RutaGaudi", and we have movement in the BD and significant delays.

In general we see the evidence with JBoss are worse than Tomcat, and it appears that page to update any impact on the performance of others that have not been touched. We believe that this should not be so.

Latest tests with Java 1.6 platform in Magnolia 3.6.1 + JBoss + MySQL

in the file prueba_081203_JBoss_jdk16_SinParametros-Tiempo Respuesta.pdf are with JBOSS with the template without parameters and we can see:
1. as in the second 200 is an update of a page different from the test and response times are multiplied by 10, but are acceptable.
2. as in the second 340 is an update to the page of the test and response times are also increasing, though less than before.

According to the evidence seems to be that whenever a page is updated on amending the entire cache of Public.

in the file prueba_081203_JBoss_jdk16-ConParametros-Tiempo Respuesta.pdf are with JBOSS with the template with parameters for the language es_ES and the language en_US, language ca_ES is the default language of the template and do not use the parameter. We can see:
1. the actual values are from the second 200, which is the starting of JBoss, this has no value.
2. times for the first column in the Catalan language (without parameters) are quite correct, similar to the previous test, but the other 2 languages are in the order of 10 times.
3. as in the second 520 is an update of a page different from the test and response times for the Catalan are multiplied by 10, nor in the other note, as are high.
4. as in the second 700 is an update to the page of the test and behaves like another page.

With the upgrade of java has earned enough respect to the previous tests, it looks like there are no jumps of 5, 8 seconds before ox if passed and randomly in all languages, and this does not happen now, all the times Now follow the same pattern.

In the previous tests can also be seen as performance is better than Tomcat with JBoss (with the same machine, same MySQL and the same values of memory).

With these and previous tests can give us a logical explanation of what is happening, whether the cache should behave like it or not, and if the JBoss we should give better results than the Tomcat.


Generated at Mon Feb 12 03:37:23 CET 2024 using Jira 9.4.2#940002-sha1:46d1a51de284217efdcb32434eab47a99af2938b.